Image sharpening and re-sampling method

Abstract

Image processing method wherein an image is re-sampled and sharpened by subjecting the image to a convolution with a kernel, the elements of which are determined by selecting a first set of kernel values so that a convolution of the image by means of a kernel with said first set of kernel values generates a sharpened image, and subsequently interpolating between the elements of this first kernel.

What is claimed is:  
     
       1. An image processing method for resampling and sharpening an image comprising: 
       generating a first convolution kernel having first kernel elements that upon convolution with said image would generate a sharpened image,  
       generating a second convolution kernel having second kernel elements wherein the generating of the second convolution kernel includes interpolating between the first kernel elements of said first convolution kernel, and  
       convolving said image be means of said second convolution kernel.  
     
     
       2. An image processing method according to claim  1  wherein said image is a medical image. 
     
     
       3. An image processing method according to claim  1  wherein the interpolating includes generating an interpolation function through the first kernel elements of said first kernel. 
     
     
       4. An image processing method according to claim  3  wherein interpolation is performed by applying cubic spline approximation of said interpolation function through the elements of said first kernel. 
     
     
       5. An image processing method according to claim  3  wherein said interpolation function has the shape of a sin(×)/× function, wherein x is variable. 
     
     
       6. An image processing method according to claim  1  wherein said kernel elements are stored in a convolution look up table. 
     
     
       7. An image processing method according to claim  1  wherein said second convolution kernel is a one-dimensional kernel and wherein said convolving said image comprises a first and second convolution with said one-dimensional kernel, and wherein said image is rotated by 90 degrees during the second convolution relative to the first convolution.